After watching the video, my first thought was we can apply to a song recommendation system.
I found a spotify_million, which contains 1 million playlists created by users on Spotify.
Each playlist represents a transaction, and the songs in the playlist are the items in the transaction. We can apply Apriori algorithm to calculate the frequent itemsets and generate association rules.
We can recommend songs based on the discovered association rules. For example, if a user likes song A and song B, we can recommend song C to them if we find a rule that says "If a user likes song A and song B, they are likely to like song C".